Citations

PURPOSE Large language models (LLMs) have demonstrated remarkable versatility in oncology applications, such as cancer staging and survival analysis. Despite their potential, ethical concerns such as data privacy breaches, bias in training data, lack of transparency, and risks associated with erroneous outputs pose significant challenges to their adoption in high-stakes oncology settings. Therefore, we aim to explore the ethical challenges associated with LLM-based applications in oncology and evaluate emerging techniques designed to address these issues. METHODS Following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses framework, a systematic review was conducted to evaluate publications related to ethical issues of LLMs in oncology across eight academic databases (eg, PubMed, Web of Science, and Embase) between January 1, 2019, and December 31, 2024. RESULTS The search retrieved 4,319 published articles, of which 65 publications were preserved and included in our analysis. We identified six prevalent ethical challenges in oncology, including trust, equity, privacy, transparency, nonmaleficence, and accountability. We then evaluated emerging technical solutions to mitigate ethical challenges and summarized evaluation metrics used to assess these solutions' effectiveness. CONCLUSION This review provides actionable recommendations for responsibly deploying LLMs in oncology, ensuring adherence to ethical guidelines, and fostering improved patient outcomes. By bridging technical and clinical perspectives, this review offers a foundational framework for advancing ethical artificial intelligence applications in oncology and highlights areas for future research.

Abstract Motivation Compositional data are frequently encountered in many disciplines, such as in next-generation sequencing experiments widely used in biomedical studies. Regression analysis with compositional data as either responses or predictors has been well studied. However, when both responses and predictors are compositional, the inventory of analysis tools is surprisingly limited, especially in the high-dimensional setting. Among the few existing methods, most of them rely on a log-ratio transformation to move compositional data from the simplex to real numbers. Yet, a serious weakness of these methods is their failure to handle the substantial fraction of zeroes observed in data collected from next-generation sequencing experiments. Results To investigate associations between two high-dimensional multi-omics compositions, we propose a composition-on-composition (COC) regression analysis method which does not require log-ratio transformations and hence can handle zeroes in the data. To account for high dimensionality, we estimate regression coefficients using a penalized estimation equation approach. Finally, inference procedures for COC regression are also proposed. Superior performance of COC is demonstrated through both comprehensive numerical simulations and case studies. Availability and implementation Source R codes to implement COC method is available at https://github.com/nrios4/COC.

Despite significant improvements in survival of patients with multiple myeloma (MM), outcomes remain heterogeneous, and a significant proportion of patients experience suboptimal outcomes. Importantly, traditional prognostic factors based on data from patients treated with older therapies no longer capture prognosis accurately in the contemporary era of novel triplet or quadruplet therapies. Therefore, risk stratification requires refinement in the context of available and investigational treatment options in routine practice and clinical trials, respectively. The current identification of high-risk MM (HRMM) in routine practice is based on the Revised International Staging System, which stratifies patients using a combination of widely available serum biomarkers and chromosomal abnormalities assessed via fluorescence in situ hybridization. In recent years, a substantial body of evidence concerning additional clinical, biological, and molecular/genomic prognostic factors has accumulated, along with new MM risk stratification tools and consensus reports. The International Myeloma Society, along with the International Myeloma Working Group, convened an Expert Panel with the primary aim of revisiting the definition of HRMM and formulating a practical and data-driven consensus definition, based on new evidence from molecular/genomic assays, updated clinical data, and contemporary risk stratification concepts. The Panel proposes the following Consensus Genomic Staging (CGS) of HRMM which relies upon the presence of at least one of these abnormalities: (1) del(17p), with a cutoff of >20% clonal fraction, and/or TP53 mutation; (2) an IgH translocation including t(4;14), t(14;16), or t(14;20) along with 1q+ and/or del(1p32); (3) monoallelic del(1p32) along with 1q+ or biallelic del(1p32); or (4) β2 microglobulin ≥5.5 mg/L with normal creatinine (<1.2 mg/dL).

Abstract Ischemic cardiomyopathy (ICM) is a leading cause of heart failure characterized by extensive remodeling of the cardiac extracellular matrix (ECM). While initially adaptive, ECM deposition following ischemic injury eventually turns maladaptive, promoting adverse cardiac remodeling. The strong link between the extent of fibrosis and adverse clinical outcomes has led to growing interest in ECM targeted therapies to prevent or reverse maladaptive cardiac remodeling in ICM; yet, the precise composition of the ECM in ICM remains poorly defined. In this study, we employed a sequential protein extraction enabled by the photocleavable surfactant Azo to enrich ECM proteins from left ventricular tissues of patients with end-stage ICM (n=16) and nonfailing donor hearts (n=16). High-resolution mass spectrometry-based quantitative proteomics identified and quantified over 6,000 unique protein groups, including 315 ECM proteins. We discovered significant upregulation of key ECM components, particularly glycoproteins, proteoglycans, collagens, and ECM regulators. Notably, LOXL1, FBLN1, and VCAN were among the most differentially expressed. Functional enrichment analyses revealed enhanced TGFβ signaling, integrin-mediated adhesion, and complement activation in ICM tissues, suggesting a feedback loop driving continued ECM deposition in the end-stage failing heart. Together, our findings provide a comprehensive proteomic landscape of ECM alterations in the end-stage ICM myocardium and identify promising molecular targets for therapeutic intervention.

Abstract Bisphosphonates (BPs) have been the major class of medicines used to treat disorders of excessive bone loss for over five decades. Recently it has been recognized that BPs may also have additional significant beneficial extra-skeletal effects. These include a reduction of all-cause mortality and of conditions commonly linked to ageing, such as cancer and cardiovascular disease. Here we show that bisphosphonates co-localize with lysosomal and endosomal organelles in non-skeletal cells and stimulate cell growth at low doses. In vivo spatial transcriptomic analysis revealed differentially expressed senescence markers in multiple organs of aged BP-treated mice, and a shift in cellular composition toward those of young counterparts. Similarly, a 5000-plex plasma proteome analysis from osteopenic patients before and after BP-treatment showed significant alterations in ∼400 proteins including GTPase regulators and markers of senescence, autophagy, apoptosis, and inflammatory responses. Furthermore, treatment with BPs protected against the onset of senescence in vitro . Proteome-wide target deconvolution using 2D thermal profiling revealed novel BP-binding targets (PHB2, ASAH1), and combined with RNA- and ATAC-seq of BP-treated cells and patient data, suggests downstream regulation of the MEF2A transcription factor within the heart. Collectively, these results indicate how BPs may beneficially modify the human plasma proteome, and directly impact multiple non-skeletal cell types through previously unidentified proteins, thereby influencing a range of pathways related to senescence and ageing.

Abstract Primary testicular diffuse large B-cell lymphoma (PT-DLBCL) is a rare and aggressive lymphoma with molecular heterogeneity not well characterize. In this study, we performed next-generation sequencing analysis for a large number of DNA and RNA samples from patients with PT-DLBCL. DNA sequencing analysis identified ≥ 3 chromosomes with copy number variations (CNVs) and microsatellite instability as prognostic biomarkers, rather than MYD88 mutations and genetic subtypes. Remarkably, targeted RNA-seq analysis in 195 patients revealed that TP53 mutations with a ≥ 40% variant allele frequency had significantly adverse prognostic impact, and that a 150-gene expression signature subdivided PT-DLBCL into two distinct clusters, termed as testicular lymphoma tumor (TLT) and microenvironment (ME) subtypes. The TLT subtype featured upregulation of genes involved in B-cell receptor signaling, cell cycle, DNA damage and repair, higher frequencies of CNVs and MYD88 mutations, elder ages, larger tumor sizes, and significantly poorer survival. Genomic microRNA profiling analysis identified significantly differentially expressed microRNAs between 113 PT-DLBCL and 180 systemic DLBCL patients, and further subdivided the PT-DLBCL cohort by microRNA signatures. The subcohort with upregulation of 16 microRNAs associated with PT-DLBCL and testicular tissue expression had significantly better survival. This study revealed characteristic genetic, gene expression, and microRNA profiles and heterogeneity in PT-DLBCL.

Introduction: Primary diffuse large B-cell lymphoma (DLBCL) of the central nervous system (CNS), traditionally referred to as primary CNS lymphoma (PCNSL), is a rare type of primary large B-cell lymphoma of immune-privileged sites with significantly poorer prognosis than systemic DLBCL. Genetic profiles of PCNSL were widely studied, whereas the tumor microenvironment has not been well characterized. Patients and Methods: A series of molecular analyses were performed using DLBCL FFPE tissues collected by the International Lymphoma Consortium. First, multiplex fluorescent immunohistochemistry (mfIHC) using antibodies for CD20, PAX5, CD3, CD4, CD8, FOXP3, CD45RO, CD68, CD56, PD-1, PD-L1, PD-L2, and CTLA-4 was performed on PCNSL tissues from 121 immunocompetent patients and de novo systemic DLBCL tissues from 405 patients. Digital quantification results for 405 patients with EBV-negative systemic DLBCL or primary testicular (PT) DLBCL have been reported previously. The preliminary mfIHC results of PCNSL will be presented. Second, mfIHC for multiple B cell biomarkers, CD3, CD4, CD8, CD68, CD163, CD11c, and CD31 were performed for 10 PCNSL validation cases, 24 PT-DLBCL and 451 systemic DLBCL tissues (unpublished). Third, RNA samples were extracted and sequenced for the targeted 1408 cancer-associated genes, successfully in 142 PCNSL cases, including 102 cases evaluated by either of these two mfIHC panels and additional 40 PCNSL cases, Results: First, in the first large mfIHC cohort, PCNSLs compared with de novo systemic nodal or extranodal DLBCL had significantly higher mean/median CD56+ cell densities whereas lower T cell densities (consistent with the immune privilege), as well as lower CD45RO+, PD-1+, and FOXP3+ percentages in T cells whereas higher CD4¯CD8¯ double-negative percentages in T cells. Prognostic analysis in the PCNSL cohort showed that high CD56+ cell densities and high CD45RO+CD8+ T cell densities were associated with significantly poorer and better patient survival, respectively, which are opposite to their prognostic effects in systemic DLBCL as previously reported. Twelve patients had very high PD-L1 expression in B cells and significantly poorer survival in the PCNSL cohort. Second, as validation, the 10 PCNSL cases in the second mfIHC study showed significantly less T cells than systemic DLBCL and PT-DLBCL cases, and low CD8 T cell densities and having low percentage of CD8 T cells adjacent to tumor cells were associated with significantly poorer survival. Immunofluorescence landscapeand spatial analysisrevealed significant differences in immune privilege between PCNSL and PT-DLBCL. Third, gene expression profiling analysis by targeted RNA-seq identified 282 significantly differentially expressed genes between PCNSL and systemic DLBCL, referred to as the PCNSL signature. Unsupervised clustering analysis divided the PCNSL cohort into two clusters, which had no significant difference in MYD88 mutation frequencies analyzed at the mRNA level. Importantly, the cluster with higher expression of many neural glial genes in the PCNSL signature had significantly poorer survival, which remained significant in multivariate analysis adjusting for clinical factors and/or treatments, two PCNSL subcohorts (as training and validation cohorts), and methotrexate-treated patients. NCAM1 (Neural Cell Adhesion Molecule 1, encoding CD56) was among the upregulated PCNSL signature genes, explaining the poorer prognosis associated with higher CD56+ cell densities in PCNSL by the mfIHC analysis. Summary: Our studies represent the largest immunophenotypic and gene expression profiling studies currently in PCNSL, a rare type of primary DLBCL of immune-privileged sites. Multiplex immunofluorescence revealed the immune landscape in PCNSL in comparison with systemic and testicular DLBCL, and identified prognostic immune biomarkers. Targeted RNA-seq identified a PCNSL gene signature enriched with neuronal-glial genes and showing significant prognostic effect in PCNSL. These novel findings in the PCNSL microenvironment warrant further studies.

Abstract Heterotopic ossification (HO), often arising in response to traumatic challenges, results from the aberrant osteochondral differentiation of mesenchymal stem cells (MSCs). Nevertheless, the impact of trauma‐induced inflammatory exposure on MSC fate determination remains ambiguous. In this study, the cellular diversity within inflammatory lesions is elucidated, comprising MSCs and several innate and adaptive immune cells. It is observed that quiescent MSCs transition into cycling MSCs, subsequently giving rise to chondrogenic (cMSC) and/or osteogenic (oMSC) lineages within the inflammatory microenvironment following muscle or tendon injuries, as revealed through single‐cell RNA sequencing (scRNA‐seq), spatial transcriptome and lineage tracing analysis. Moreover, these investigations demonstrate that neutrophils and natural killer (NK) cells enhance transition of quiescent MSCs into cycling MSCs, which is also controlled by M1 macrophages, a subpopulation of macrophages can also stimulate cMSC and oMSC production from cycling MSCs. Additionally, M2 macrophages, CD4 + and CD8 + T lymphocytes are found to promote chondrogenesis. Further analysis demonstrates that immune cells promotes the activation of signaling transducers and activators of transcription (STAT) pathway and phosphoinositide 3 (PI3K)/protein kinase B (AKT) pathway in MSC proliferation and osteochondral progenitors’ production, respectively. These findings highlight the dynamics of MSC fate within the inflammatory lesion and unveil the molecular landscape of osteoimmunological interactions, which holds promise for advancing HO treatment.

AbstractObjectiveThis study investigated the comparative performance of ear, nose, and throat (ENT) physicians in correctly detecting ear abnormalities when reviewing digital otoscopy imaging using 3 different visualization methods, including computer‐assisted composite images called “SelectStitch,” single video frame “Still” images, and video clips. The study also explored clinicians' diagnostic confidence levels and the time to make a diagnosis.Study DesignClinician diagnostic reader study.SettingOnline diagnostic survey of ENT physicians.MethodsNine ENT physicians reviewed digital otoscopy examinations from 86 ears with various diagnoses (normal, perforation, retraction, middle ear effusion, tympanosclerosis). Otoscopy examinations used artificial‐intelligence (AI)‐based computer‐aided composite image generation from a video clip (SelectStitch), manually selected best still frame from a video clip (Still), or the entire video clip. Statistical analyses included comparisons of ability to detect correct diagnosis, confidence levels, and diagnosis times.ResultsThe ENT physicians' ability to detect ear abnormalities (33.2%‐68.7%) varied depending on the pathologies. SelectStitch and Still images were not statistically different in detecting abnormalities (P > .50), but both were different from Video (P < .01). However, the performance improvement observed with Videos came at the cost of significantly longer time to determining the diagnosis. The level of confidence in the diagnosis was positively associated with correct diagnoses, but varied by particular pathology.ConclusionThis study explores the potential of computer‐assisted techniques like SelectStitch in enhancing otoscopic diagnoses and time‐saving, which could benefit telemedicine settings. Comparable performance between computer‐generated and manually selected images suggests the potential of AI algorithms for otoscopy applications.

Importance The clinical effects of risankizumab (a monoclonal antibody that selectively targets the p19 subunit of IL-23) for the treatment of ulcerative colitis are unknown. Objective To evaluate the efficacy and safety of risankizumab when administered as an induction and a maintenance therapy for patients with ulcerative colitis. Design, Setting, and Participants Two phase 3 randomized clinical trials were conducted. The induction trial was conducted at 261 clinical centers (in 41 countries) and enrolled 977 patients from November 5, 2020, to August 4, 2022 (final follow-up on May 16, 2023). The maintenance trial was conducted at 238 clinical centers (in 37 countries) and enrolled 754 patients from August 28, 2018, to March 30, 2022 (final follow-up on April 11, 2023). Eligible patients had moderately to severely active ulcerative colitis; a history of intolerance or inadequate response to 1 or more conventional therapies, advanced therapies, or both types of therapies; and no prior exposure to risankizumab. Interventions For the induction trial, patients were randomized 2:1 to receive 1200 mg of risankizumab or placebo administered intravenously at weeks 0, 4, and 8. For the maintenance trial, patients with a clinical response (determined using the adapted Mayo score) after intravenous treatment with risankizumab were randomized 1:1:1 to receive subcutaneous treatment with 180 mg or 360 mg of risankizumab or placebo (no longer receiving risankizumab) every 8 weeks for 52 weeks. Main Outcomes and Measures The primary outcome was clinical remission (stool frequency score ≤1 and not greater than baseline, rectal bleeding score of 0, and endoscopic subscore ≤1 without friability) at week 12 for the induction trial and at week 52 for the maintenance trial. Results Among the 975 patients analyzed in the induction trial (aged 42.1 [SD, 13.8] years; 586/973 [60.1%] were male; and 677 [69.6%] were White), the clinical remission rates at week 12 were 132/650 (20.3%) for 1200 mg of risankizumab and 20/325 (6.2%) for placebo (adjusted between-group difference, 14.0% [95% CI, 10.0%-18.0%], P &amp;lt; .001). Among the 548 patients analyzed in the maintenance trial (aged 40.9 [SD, 14.0] years; 313 [57.1%] were male; and 407 [74.3%] were White), the clinical remission rates at week 52 were 72/179 (40.2%) for 180 mg of risankizumab, 70/186 (37.6%) for 360 mg of risankizumab, and 46/183 (25.1%) for placebo (adjusted between-group difference for 180 mg of risankizumab vs placebo, 16.3% [97.5% CI, 6.1%-26.6%], P &amp;lt; .001; adjusted between-group difference for 360 mg of risankizumab vs placebo, 14.2% [97.5% CI, 4.0%-24.5%], P = .002). No new safety risks were detected in the treatment groups. Conclusion and Relevance Compared with placebo, risankizumab improved clinical remission rates in an induction trial and in a maintenance trial for patients with moderately to severely active ulcerative colitis. Further study is needed to identify benefits beyond the 52-week follow-up. Trial Registration ClinicalTrials.gov Identifiers: NCT03398148 and NCT03398135